Introduction

Reflective processes are linked to mental health and wellbeing for both clinicians and clients, and numerous constructs have emerged in the psychotherapy literature to describe this phenomenon. Mentalization was defined by Fonagy (2002) as the ability to be aware of one’s internal mental states, and to reflect on the thoughts, feelings and intentions of others. Emerging from the psychoanalytic and attachment theory traditions, this capacity is viewed as essential to the process of making meaning in relation to others and the self, and was later operationalised as reflective function (RF; Fonagy & Bateman, 2007). RF is associated with early caregiving experiences in that the presence of a caring adult who treats the infant as an intentional agent and supports them to name and describe their emotions helps to develop this capacity (Nazzaro et al., 2017). Higher scores for RF are associated with mental wellbeing (Fonagy, 2002; Katznelson, 2014) and are protective against psychopathology (Berthelot et al., 2019). Low scores in RF have been linked to severe psychological problems including personality disorders (Fonagy & Bateman, 2007; Newbury-Helps et al., 2017), eating disorders (Cucchi et al., 2018; Sacchetti et al., 2019), depressive disorders (Scandurra et al., 2020) and psychosis (MaBeth et al., 2011). RF is also positively associated with psychotherapy alliance and outcomes, with higher RF in therapists associated with greater therapy effectiveness and therapist wellbeing (Brugnera et al., 2020; de la Cerda et al., 2019). Such findings have contributed to the perceived value of promoting reflective practice in training of mental health professionals (Reynolds, 2011).

Metacognition appears to have some definitional overlap with mentalization despite theoretical and conceptual differences (Ridenour et al, 2019) and is defined as the ability to think about one’s own mental state and the mental states of others (Semerari et al., 2003). This construct involves three sub-domains, understanding one’s own mind, understanding other’s minds and mastery. These abilities are described as allowing for the integration of experiences as influenced by the motives of others, to produce a stable sense of self (Dimaggio et al., 2007). Impairments have been described in metacognition with psychotic disorders (Lysaker et al., 2013) and personality disorders (Bourke & Grenyer, 2017).

A variety of alternative descriptions have been used by researchers who have taken up discursive methodologies to put forward an intersubjective account of reflective processes as they occur in conversation. Wodak (1981) described the development of ‘linguistic awareness’ or a client’s contemplation of their own speech. This awareness and articulation of one’s own internal experiences is understood to support the development of agency and social functioning (Rennie, 2004).

Psychotherapy models such as Mentalization based therapy (MBT; Brent et al., 2014), Metacognitive reflection and insight therapy (MERIT; James et al., 2018; Lysaker et al., 2019) and the Reflecting team approach (Andersen, 1987) each take up these ideas in therapy to encourage reflection in clients. Despite the differences in epistemological perspective, the process of, or ability to, reflect on oneself or others is widely accepted as valuable in psychotherapy and relevant to psychotherapy process research (Goodman et al., 2016). This scoping review aims to provide an integrated analysis of studies purporting to measure and describe the process of reflection during psychotherapy.

Method

Design

Given the complex and heterogeneous nature of the studies reviewed here we chose a systematic review with a scoping aspect (Peters et al., 2015). Scoping reviews are used to clarify theoretical and working definitions or the boundaries of a concept within a given field and can also present information on how research has been conducted. The protocol for this review was developed a priori with a team of researchers with experience relevant to the aims of this study. The following questions were the basis for this review: (1) How is reflection understood from varying theoretical perspectives in psychotherapy process research? (2) How do these understandings influence the measurement of this phenomenon? (3) What can be learned from integrating these understandings in relation to psychotherapy research and practice? Given our aim was to focus on psychotherapy process research we have prioritised studies using psychotherapy recordings to track this phenomenon.

Search Strategy

We used COVIDENCE software to support review, quality evaluation and extraction of data for papers identified in this study. As this review was focussed on both qualitative and quantitative studies, the SPIDER protocol was used to develop inclusion criteria (Cooke et al., 2012). For this study ‘reflective process’ was operationalised as elements of conversations in which participants were able to look upon and articulate their own thoughts and inner responses, or to express their understanding or their curiosity about the inner responses of others. While a wide variety of terms and constructs with some similarity emerged in the literature, our intention was to focus on studies that describe measurement of in situ during therapy conversations. This rendered three categories in the literature: Mentalization, metacognition, and studies adopting a discursive methodology. Studies that were not in English, did not use therapy transcripts or did not focus on reflective processes were excluded.

Screening Procedure

An initial search was conducted followed by an analysis of relevant article title, abstract and text keywords to produce key search terms for a more targeted and comprehensive search. A second search was undertaken using all identified relevant keywords and index terms. Following the database search the lead author conducted an initial screen to remove clearly irrelevant articles from the records (3367). We conducted a final extensive search in MEDLINE, PsychINFO, Web of Science and SCOPUS databases in July 2022. In order to locate recent findings, this search was limited to articles published after 1995. Reference lists from selected articles were reviewed and four additional articles were identified. Following exclusion, 345 studies were imported into COVIDENCE software for screening. Team members reviewed the lead author’s screened articles and for some articles team members with specific areas of expertise (e.g. in linguistic methods) provided in depth review to determine study inclusion. Assessment of study quality and risk of bias was conducted based on the Cochrane Handbook for Systematic Reviews (Flemming et al., 2018) (Fig. 1).

Fig. 1
figure 1

PRISMA flow diagram

Full size image

Results

We identified 22 studies with a focus on reflective processes in psychotherapy conversations suitable for this study. Studies ranged across a number of theoretical and epistemological perspectives. Methodological approaches were associated with the researchers’ epistemological position and conceptualisation of the reflective process. Most of the studies identified were small sample or case study designs which aimed to describe process factors in detail or to determine the utility of a particular method for capturing reflective processes during psychotherapy.

Reflective Process Terms and Definitions

A number of studies included in this review employed Fonagy’s construct of mentalization (Fonagy, 2002). One study (Hörz-Sagstetter et al., 2003) noted the similarity to aspects of social cognition, mindfulness and psychological mindedness. Another study utilised the construct of mentalization but posed a definition that included interactional processes rather than a distinct internalised ability (Keselman et al., 2018). These authors further expanded on the mentalization concept by differentiating between explicit mentalizing, identified through the client’s reflective verbal response, and implicit mentalizing, which could be inferred by the client’s apparent failure to respond to the therapist’s invitation to reflect. This implicit mentalization was noted to achieve social aims such as controlling the content of the session, or equalising power in the client therapist relationship (Keselman et al., 2018).

Metacognition was a term used in several studies and was based on Semerari’s definition (2003). Included studies employing this definition focussed on the therapy of clients with psychotic and personality disorders, which aimed to promote metacognition (Arnon-Ribenfeld et al., 2018; Lysaker & Gumley, 2010) while another study that situated itself within metacognition research evaluated an Emotion Focussed Therapy program (Carcione et al., 2008).

Other studies adopted terms associated with the Narrative Process Coding System (NPCS) and corresponding theory. Angus et al. (2013) describe “reflective storytelling” as a reflexive sequence of talk between the client and therapist which involves describing a particular event in terms of the thoughts, feelings, beliefs or intentions of others. The authors identify themselves as adopting a dialectical constructivist viewpoint. This marks a shift from the process of interest being located in the client as a particular attribute or skill, to a discursive process occurring between therapist and client. Two additional studies in our review also adopt this viewpoint to identify interactional process markers in individual and family therapies (Boritz et al., 2014; Laitila et al., 2001). A third method known as the Innovative Moments Coding System was employed to compare good and poor outcome case studies (Cunha et al., 2013). Innovative moments were defined as reflections on the self and others, which allowed for the development of a new narrative.

Most studies adopting qualitative or linguistic methods in this review tended to utilise descriptive terms to describe the reflective process. One study simply used “self-reflective talk” which was defined as a social display of cognitive processing (Logren et al., 2017). Esposito et al. (2019) examined two processes in therapy conversations: “reflection” and “reflexivity”. The authors define reflection as the psychological process allowing an individual to “access a new understanding” by which the subject can think about relational processes relating to themselves. Reflexivity is described as occurring whenever a person can understand the effects of their actions in an intersubjective context. Strong (2006) used the term reflection, which was defined as an opportunity to examine taken-for-granted aspects of life and was likened to meditation. Strong further defined this as a dialogic “accomplishment” between the therapist and client. Two other studies took up constructs associated with the particular therapy model participants engaged in the study. Similarly, Henderson-Brooks (2010) uses the term “reflective consciousness” which is defined as a language based symbol of the self in relation to others (Meares, 1998). Rennie used the term “client reflexivity” to describe a self-evaluation that supports agency (Rennie, 2000).

Methodological Approaches and Process Outcomes

Most studies in this review attempted to apply previously developed clinician rating scales to therapy transcripts in order to gauge the extent of RF displayed. Babl and colleagues found that RF improved over time during Cognitive Behavioural Therapy (N = 37), and that this improvement was associated with decreases in depressive symptomatology (Babl et al., 2022). Talia et al. (2019) compared the in-session version of the RF Scale to the Exploring Scale from the Patient Attachment Coding System (PACS). This study analysed sessions from 160 patients across a variety of psychotherapy modalities. The in-session RF Scale is a variation of the RF scale typically used to rate responses during the Adult Attachment Interview (AAI; Taubner et al., 2013). Researchers are trained to review sections (150 word segments) of content that are then rated on a scale between − 1 and 9 based on the presence of absence of verbalisations related to mental state. The PACS Exploring scale is rated based on the frequency and intensity with which nine discourse markers appear in a transcript. These markers are grouped under three subscales: “Self-asserting,” which may be viewed as a measure of patients’ expressed agency, “Affective sharing” conveying emotional experience and, “Autonomous reflection” relating to the internal experience of oneself or other people. These researchers found that therapy participants with higher ratings on the RF Scale also showed high in-session ratings for these three features, leading them to conclude that these scales are useful measurement tools for process research. Despite extensive data (160 patients, most with over 20 sessions), they did not find changes in autonomous reflection measures during the course of therapy. This is counter to expectations and, given the large size of the study and questions assumptions regarding both expected improvements related to therapy and how these changes are measured.

A similar study but with a much smaller number (n = 2) described qualitative markers of RF as attempts by the client to understand themselves and others in terms of mental states (Hörz-Sagstetter et al., 2003). This study attempted to track improvements in scores of the RF scale over 300 h of therapy, and included a measure of therapist actions (psychotherapy Q-set). They found that in-session RF (tracked across 150 word blocks) fluctuated during the session, with a trend towards increased RF in sessions beyond 240 therapy hours. These fluctuations appeared to be mediated by therapist’s invitations to reflect and overall scores were related to baseline measures of RF conducted using the AAI.

Another study adopting the construct of RF employed the Experiencing Scale for a time series analysis comparing early sessions (1–10) to later (50–60) for three therapy participants (Brockmann et al., 2018). The Experiencing scale is a rating system with seven levels ranging from client talk about events, or people, to evidence of “reflective self-awareness” (Klein et al., 1986). The results of this study were again unexpected: one participant showed no change over time, and the other two indicated a reduction in RF based on results of this measure on in-session transcripts. The authors noted an inverse correlation between Therapeutic Alliance (rated using the Therapeutic Alliance Scale at each session) and reflective self-awareness. They also found a correlation between therapist intervention and measures of reflective self-awareness but were unable to make conclusions regarding directionality. Authors suggest a reciprocal causal process between therapist intervention and reflective self-awareness may explain these results.

Many of the articles utilising the metacognition construct used the Metacognition Assessment Scale (Leonhardt et al., 2006; Semerari et al., 2003). The abbreviated version adopted by Arnon-Ribenfeld et al. (2018) is a coding system that includes four subscales: self-reflectivity, understanding the mind of the other, decentration, and mastery. Differences between early, mid and final sessions were compared using the Reliable Change Index in two patients diagnosed with schizophrenia. While two subscales (understanding others’ minds and mastery) showed significant differences, no change was evident for self-reflectivity. Lysaker and Gumley (2010) reviewed therapy transcripts to explore the long term therapy of a client diagnosed with schizophrenia. The authors aimed to develop a model for how psychotherapy might promote metacognition in people with psychosis. They identified three themes: unbearable affect (experiences of pain and rejection), therapist activity (the use of clarifying questions or empathy to attune to the client) and evolving reflective awareness (an interactive process of the therapist attending to painful affect and the client’s attempts to make sense of complex emotions).

The NPCS (Angus et al., 1999) was taken up by a number of authors adopting a constructivist perspective. This system developed as a method for assessing client change processes during psychotherapy. Therapy transcripts are divided according to content shifts, and then identified as one of three dialectical constructivist process modes; external sequences that include the disclosure of personal stories; internal sequences include descriptions of embodied feelings and emotions; and reflexive sequences entailing recursive questioning, meaning-making processes and reflection on core beliefs and personal stories (Angus et al., 2017). Banham and Schweitzer (2015) compared three clients with good and three with poor outcomes using this approach. They found clients with good outcomes had a higher frequency of reflexive narrative modes. Using the Narrative-emotion Process coding system (NEPCS) Boritz and colleagues compared early, middle and late sessions from cognitive-behavioural, emotion focussed and client centred therapies (Boritz et al., 2014). They described greater instances of change markers in later sessions of patients who reported recovery from depression following the treatment phase. The two markers noted as showing a significant outcome effect were inchoate storytelling (a segment marked by fragmented or disrupted speech as the client struggled to articulate experience) and discovery storytelling, which entails an experiential description alongside a “reflective component or interpretive analysis of the subjective feeling or event” (p. 597). In line with the theoretical position underpinning this model of therapy, the researchers’ methodology positions the interactive quality of the dialogue between the client and therapist as the major variable to be considered (Angus & Hardtke, 1994). Laitila and colleagues expanded this understanding by studying Narrative Process modes in Family therapy (Laitila et al., 2001, 2005). These studies took up a particular focus on co-created narratives between therapist and family members, and described the emergence of a shared reflective narrative between family members as a form of “reflective expertise (2001, p. 321).

A small group of studies in this review adopted discursive methodologies. Kesselman and colleagues took a conversational analytic approach to mentalization as it occurs in therapy sessions (Keselman et al., 2018). Conversational analysis understands language in terms of what it achieves in social interaction (Pain, 2009; Sacks & Jefferson, 1995). In this study, authors were particularly interested in how clients responded to therapists’ invitations to mentalize. Interactions in which the client refrains from explicit mentalization were understood not as a deficit but rather as implicit mentalization, which served a particular function in the exchange. Emerging themes were that rejections of these reflection invitations were attempts by the client to gain control of the session content, to protect themselves from challenging subject matter, or to engage in confrontation to express honesty (Keselman et al., 2018). This study challenges the unwritten assumption that the absence of phrases or words indicating reflection indicates impairment, and instead suggests more complicated social interactions are present for these clients. Esposito et al. s’ (2021) discourse analytic approach used a word set associated with reflection such as “think” and “assume” to track reflective and reflexive verbalisations. This study followed seven group therapy participants and noted an increase in the use of these words over the course of treatment.

The remaining studies in this review were qualitative explorations of the reflective process. Both Logren et al. (2017) and Strong (2006) took an explicit social constructivist position, adopting Conversational Analysis as their approach. Both studies report viewing reflective processes as ‘joint social action’ and explored how this emerges during psychotherapy. Logren et al (2017) explored talk during group counselling sessions, and reported reflections emerging following a therapist’s expression of curiosity, or another participant’s disclosure. Strong (2006) viewed reflection as a dialogic accomplishment, involving at least two persons in dialogue. He noted in his analysis of 21 sessions that reflective content occurred when therapists invited clients to take up a position on something said, or to enquire about how they came to hold a particular understanding. Rennie (2000) described a hermeneutical grounded theory approach, which focused on researcher reflexivity in examining transcripts of both therapy sessions and interpersonal process recall sessions following them. Interpersonal process recall invites the client to comment on interactions as they have occurred by viewing session recordings. This study found that the client’s reflexivity was associated with the exertion of control over the session content. Henderson-Brooks describes a partly quantitative discourse analytic study drawing on Systemic Functional Linguistics (Halliday, 2003) to track changes over time for participants in Conversational Therapy (Henderson-Brooks, 2010). Using transitivity analysis (Halliday, 2003), this study showed that later sessions with seven patients diagnosed with Borderline Personality Disorder displayed an increase in the rate of mental clauses produced per session (i.e. sentence segments in which meaning centres around verbs such as thinking or imagining), and this was seen as evidence of a trend towards using reflective narratives.

Practice Implications

While some studies included in this review were designed to assess the suitability of a particular methodology for tracking reflective processes during psychotherapy, most aimed to study psychotherapy process. These study designs followed the same patient at early, middle and late stages of therapy. Other studies presented unexpected results such as RF remaining stable, or decreasing over therapy sessions (Brockmann et al., 2018). In another study fluctuations in RF were noted to be related to baseline measures of RF and it was difficult to determine the effect of therapy on measurements of RF (Hörz-Sagstetter et al., 2003). A study seeking to track change in metacognition during therapy did not find significant differences in “self-reflectivity” (Arnon-Ribenfeld et al., 2018).

Of note was the shift in conceptualisation proposed by Keselman et al. (2018). They conducted conversational analysis of transcript segments containing therapist invitations to reflect, and argued that mentalization is an interactive phenomenon whereby clients may choose not to take up invitations to reflect, in order to carry out a particular social function. Rennie (2000) adopted a similar position noting that clients exerted control over session content in order to elicit particular responses from the therapist. The client’s resistance to avenues of exploration can be understood as a display of reflexivity and agency.

Narrative process coding studies compared the therapy talk of good and poor outcome cases (Angus et al., 2013; Boritz et al., 2014; Laitila et al., 2001) found a higher frequency of ‘reflexive narratives’ in good outcome cases but no significant differences across therapy types (Angus et al., 2013). Two studies seeking to explore the emergence of reflection suggest that reflecting talk is associated with therapists who assert the client’s expertise, and a fellow group therapy participant’s disclosure (Logren et al., 2017; Strong, 2006). Another investigation using the Psychotherapy Process Q-set found that RF (measured by rating the RF Scale) was linked to therapist actions such as those indicating sensitivity to the emotions of the client, and raising emotions that may be difficult for the client (Karlsson & Kermott, 2006).

Discussion

Studies included in this review used talk in therapy conversations to observe reflective processes. Within the literature two main constructs were dominant: RF (mentalization) and metacognition. Studies taking up a constructivist epistemology incorporated terms such as self-reflective talk, and reflective self-awareness. Most studies in this review aimed to assess reflective processes as a psychotherapy process factor. A notable finding of the review was that some studies, including one large-scale study, did not show the expected improvement in reflective functioning over time. Some researchers suggested RF may be a fixed state, and predictive of outcome rather than causal (Karlsson & Kermott, 2006) while others concluded that the absence of explicit mentalizing does not indicate the absence of this ability (Keselman et al., 2018). This review was complicated by the variety of definitions, however, areas of overlap may offer opportunities for integration that could guide future research.

Each study took up the assumption that words are indicators of a reflective process and a useful tool for exploring this process. Methodologies are linked not just to these constructs but also to the culture and epistemological perspective of the researcher (Parker, 2015; Potter, 2003). Individualism has exerted significant influence over the way we understand human experience (Henderson, 2008) and the process of mental health recovery. Researchers adopting a positivist and individualist perspective make the additional assumption that reflective talk represents an internal ability or trait that we can learn about by studying the individual. The developers of the constructs of mentalization and metacognition assume that people with mental health concerns are lacking in this ability, and that therapy should address this impairment. Lived experience researchers have criticised the deficit model of psychopathology and the often-disempowering language associated with it (Bergström et al., 2019; Bradfield & Knight, 2008; White, 2017). These criticisms also illuminate the way in which such descriptions obscure relational context, agency of those in distress, and resistance to manifestations of power.

A number of studies in this review described a conversational analytic approach to reflective processes that occur during therapy (Keselman et al., 2018; Strong, 2006). Methodologies such as these shift perspectives by suggesting that clients may be choosing not to accept therapist invitations to mentalize in order to achieve a social function (Keselman et al., 2018). Similarly, Rennie (2000) has suggested that clients negotiate the focus of treatment and use both deference (Rennie, 1994) and storytelling (Rennie, 1994) to elicit particular responses from therapists. This perspective represents a shift away from the deficit model to consideration of the various social actions that a client’s responses might achieve. Theoretical understandings such as “relational consciousness” (Bird, 2004) invite the consideration of these processes as between people, rather than within them. John Burnham has used the term ‘relational reflexivity’ to refer to the collaborative practices through which therapists and clients explicitly engage one another in coordinating their resources (Burnham, 2018). The results of this review suggest the value of integrating varied epistemological perspectives on the reflecting process rather than considering these conceptualisations mutually exclusive. Viewing this process as relational and potentially mutually transformative (Anderson, 2012) has the potential to add to our understanding of how to improve therapy outcomes.

Limitations

Our initial search returned an extraordinarily high number of references due to the nature of terms such as “reflection”, which are used in a variety of contexts. It is possible that we missed relevant papers describing the same or similar phenomenon. Many of the studies in this review were case studies or used a small number of participants, and attempts to code for reflective processes were often poorly described.

Conclusion

This review aimed to integrate varying perspectives on a process considered a potential common factor related to effectiveness in psychotherapy (Wampold et al., 2007). How we understand this process has significant implications for psychotherapy practice and training. Viewing problems and potential solutions as relational or systemic is not new in the family therapy field (Dallos & Draper, 2010; Donovan, 2015) and critiques of individualism in mental health care are increasing (Henderson, 2008), however our results suggest further research may contribute to understanding this process from a relational perspective. To do this a clear description of replicable analysis of intersubjective therapeutic talk will be important, along with uncovering how such talk supports recovery.